cc: Keith Briffa <k.briffaatXYZxyz.ac.uk>

date: Thu, 23 Aug 2007 13:28:04 -0700 (PDT)

from: "David M. Ritson" <dmratXYZxyzc.stanford.edu>

subject: Re: RCS

to: Tim Osborn <t.osbornatXYZxyz.ac.uk>

<x-flowed>

Dear Tim and keith,

I will await your longer comments with interest. I have quickly read Melvins

thesis. Currently your reply reinforces the need for specific detailed

algorithms as to exactly what approach(es) you and others are actually taking.

The fact that there are many variants on possible approaches makes it much more

essential that the actually used choices are clearly delineated. For instance

my preliminary analysis code permits solutions by iterative relaxations to

a final solution. I noted that Melvin sometimes used this, but prior to that

I had thought all your results (and those of others) were based on a single

iteration, and accordingly believed that simulations of the extant results

should be based on a single iteration?

My preliminary reading of Melvins thesis is that he had noted the effects of

trees relative to age distribution. However you can only use extant

measurements and they use based on far from optimal age distributions. Prior to

his thesis there seems little mention of this potential systematic bias

which with most currently available data is certainly non-negligible.

Behind my questions is the idea that the current RCS results on millenial

change did not avoid the uncertainties of age-growth extrapolations back in

time associated with non-RCS. My gut feeling is that for currently available

data RCS should be used to set bounds not `exact' values. I would

propose that the first hundred years or so of juvenile grouth be cut out

as variable and subject to large uncertainties. That an `upper' bound be set by

assuming subsequent constant ring-width growth and a lower bound be set by

assuming subsequent growth with constant ring-area (not width) age-increase.

Obviously one can't sugest/document `improvements' until one knows what one

purports to improve on.

Cheers

Dave

On Fri, 17 Aug 2007, Tim Osborn wrote:

> Dear Dave,

>

> I'm leaving it to Keith to respond in detail to you about the RCS questions,

> but since he is on holiday at the moment I thought I should at least make a

> brief answer.

>

> Regarding our paper, our purpose was to try a different method of combining

> proxy records and see if that made a difference to the conclusions, and so we

> deliberately wanted to avoid changing, at the same time, the proxy records

> themselves (e.g. by standardising them differently) because changing two

> things at once would have confused the issue. This doesn't mean that we are

> completely satisfied with the way that some, or indeed all, of the proxy

> records have themselves been processed/constructed, but that is for separate

> work -- which we are now doing with respect to RCS

>

> I'll leave it to Keith to respond about the availability, or not, of a

> mathematical algorithm for the RCS method. It can be implemented in a number

> of ways, probably, and so this may be difficult (e.g. I referred to Esper's

> version as "global" RCS because he used the same expected age-growth curve

> across widely distributed sites, I believe, which is somewhat unusual).

>

> We have been aware for some years that the RCS method does not recover an

> unbiased representation of the input climate signal in certain situations,

> and, although these can be partly mitigated by combining sub-fossil and

> living material, there are particular problems near the ends of the record.

> Tom Melvin, one of Keith's PhD students, completed his thesis on this topic a

> few years back and much of his work (identifying RCS biases and testing

> potential solutions) is nearing publication now.

>

> It is, however, true to say that the RCS method, when implemented using

> samples from a number of epochs (i.e. not just modern living material) does

> not suffer from the "segment length curse" that methods which detrend each

> individual tree sample do suffer from, although, as noted above, the

> resulting chronology may suffer from other biases.

>

> That's all I'll say for now until Keith is back from vacation,

>

> Tim

>

> At 03:10 30/07/2007, David M. Ritson wrote:

>

>> Dear Tim,

>>

>> Thanks for your clarifications. However my underlying concern was to find a

>> detailed account and defense of the RCS method ("global RCS" processing).

>> I may well have missed something but was not helped by your references.

>> in this connection. I have already , as carefully as possible, read

>> the extant RCS literature starting from Fritts 1976 and Briffa's 1992 and

>> proceeding to the present.

>>

>> The claim that RCS methodology avoids the problems

>> of `the segment curse' appears to rest on an

>> (otherwise unsupported(?) claim in Cook et al 1995 that " ,,, the

>> cross-dated annual changes in ring-width between trees due to

>> climate are forced out of alignment and effectively averaged out in the

>> creation of the mean regional curve."? If Cook et al's above statement was

>> a valid approximation, RCS would indeed

>> avoid the `segment curse'. I have taken a preliminary look at this using

>> simulated data focussed on cases with sufficient data to

>> make systematics the primary final error-source.

>> In the presence of systematic trends in paleo forcing history and

>> sample-depth the Cook cancellations are less than perfect. This is of

>> course

>> preliminary. I have no desire to reinvent the wheel, and particularly not

>> to replace it with a square wheel, and may not be doing what you guys do,

>> or

>> you may have already looked at this in depth.

>> In my past life as a particle physicist, prior to analysis we would

>> write-out a detailed mathematical prescription for processing a set of

>> data. Such an algorithm was sufficient so that it could (if wished) be

>> handed over to a soft-ware programmer and he/she could then, even without

>> knowledge of the field, provide clean working code. I presume

>> that you follow this procedure. Your RCS mathematical algorithm is what I

>> am looking for, not its code implementation.

>>

>> I would be grateful if you could let me have this and of

>> course any revelant work of you or others in this area. Hopefully this is

>> not a burdensome request,

>>

>> Cheers

>>

>> Dave

>>

>>

>>

>>

>>

>>

>>

>

> Dr Timothy J Osborn, Academic Fellow

> Climatic Research Unit

> School of Environmental Sciences

> University of East Anglia

> Norwich NR4 7TJ, UK

>

> e-mail: t.osbornatXYZxyz.ac.uk

> phone: +44 1603 592089

> fax: +44 1603 507784

> web: http://www.cru.uea.ac.uk/~timo/

> sunclock: http://www.cru.uea.ac.uk/~timo/sunclock.htm

>

>

>

</x-flowed>

## No comments:

## Post a Comment